Narrative Processing Pipeline Configuration

This pipeline configuration tool is designed to facilitate researchers working with naturalistic stimuli such as movies, video clips, audio recordings, and text-based narratives.
The system provides a flexible platform to process complex narrative stimuli and data with the support of visualization, open source packages, and AI-tools.
Human raters are optional but encouraged to check the outputs. Human edits made to each processing step's outputs can be separately saved, automatically labeled by user name.

Main Functions Available:

Audio Transcription: Convert audio files (story and recall audio) to text using speech recognition models
Event Segmentation: Segment story transcripts into discrete events based on semantic boundaries
Text Correction: Apply spelling and grammar corrections to recall texts while preserving original structure
Text Parsing: Parse corrected recall texts into fine-grained segments at clause boundaries
Event Matching: Match recall segments to corresponding story events using semantic similarity

How to Configure the Processing Pipeline:

Click on any step in the "Available Steps" palette on the left to view detailed information about its function, inputs, parameters, and outputs. Drag steps from the palette to the "Pipeline Flow" canvas in the center to build your processing pipeline. You can reorder steps, configure input/output paths for each step, and preview how your pipeline will appear on the main dashboard. Once configured, click "Confirm Pipeline" to save your configuration.

Available Steps

audioTranscribe (story)

audioTranscribe (recall)

eventSegment

sentenceCorrect

textParsing

textMatching

causalRating

📋 Getting Started

Add steps: Click + Add on any step, or drag its header into the Pipeline Flow area on the right.

Reorder Steps: Once steps are in the pipeline, you can drag and drop them to swap positions and change the processing sequence.

Pipeline Flow

Rater name

Used so any output files are saved under your name. Roll the dice for a random name.

↓

Drag a step here or click + Add

Processing Panel Preview

Add steps to see preview

💡 Step Sequence Matters

Independent Steps: Some steps can be run in any order because they don't depend on each other. For example, transcribing story audio files and recall audio files can be done in any sequence.

Dependent Steps: Other steps must follow a specific sequence because one step's input depends on another step's output.

Example: textMatching requires the output from textParsing as its input. Place textParsing before textMatching in your pipeline.

This notification will disappear when you interact with the pipeline.

Settings

Choose folder

Narrative Processing Pipeline Configuration